ARTICLE DOI: 10.1038/s41467-017-00258-4
OPEN
Stochastic priming and spatial cues orchestrate heterogeneous clonal contribution to mouse pancreas organogenesis Hjalte List Larsen1, Laura Martín-Coll1, Alexander Valentin Nielsen2, Christopher V. E. Wright3, Ala Trusina2, Yung Hae Kim1 & Anne Grapin-Botton 1
Spatiotemporal balancing of cellular proliferation and differentiation is crucial for postnatal tissue homoeostasis and organogenesis. During embryonic development, pancreatic progenitors simultaneously proliferate and differentiate into the endocrine, ductal and acinar lineages. Using in vivo clonal analysis in the founder population of the pancreas here we reveal highly heterogeneous contribution of single progenitors to organ formation. While some progenitors are bona fide multipotent and contribute progeny to all major pancreatic cell lineages, we also identify numerous unipotent endocrine and ducto-endocrine bipotent clones. Single-cell transcriptional profiling at E9.5 reveals that endocrine-committed cells are molecularly distinct, whereas multipotent and bipotent progenitors do not exhibit different expression profiles. Clone size and composition support a probabilistic model of cell fate allocation and in silico simulations predict a transient wave of acinar differentiation around E11.5, while endocrine differentiation is proportionally decreased. Increased proliferative capacity of outer progenitors is further proposed to impact clonal expansion.
1 DanStem, University of Copenhagen, 3B Blegdamsvej, DK-2200 Copenhagen N, Denmark. 2 Niels Bohr Institute, University of Copenhagen, 17 Blegdamsvej, DK-2200 Copenhagen N, Denmark. 3 Department of Cell & Developmental Biology, Vanderbilt University, Nashville, TN 37232-0494, USA. Correspondence and requests for materials should be addressed to Y.H.K. (email: [email protected]) or to A.G.-B. (email: [email protected])
NATURE COMMUNICATIONS | 8: 605
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
1
ARTICLE
D
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
efining the rules governing embryonic organ development and postnatal tissue homoeostasis is essential for understanding disease pathology and for the generation of functional cell types for regenerative medicine purposes. Seminal studies have demonstrated how rapidly proliferating postnatal tissues such as the skin and the intestine are homeostatically maintained by equipotent stem cells undergoing seemingly stochastic cell fates choices by neutral competition for limited niche signals1–4. In contrast to postnatal tissue homoeostasis, embryonic development of most organs occurs at a state of system disequilibrium, as a population of progenitors expands while simultaneously giving rise to differentiating progeny. Although optimality in the design of strategies ensuring rapid organ development has been proposed5, little is known regarding how global embryonic organogenesis is orchestrated when deconstructed into clonal units originating from single progenitors at the onset of organ bud formation. Studies of retinal development have provided compelling evidence for a stochastic process of cell fate choices using both in vitro6 and in vivo approaches7. However, a deterministic model of embryonic neocortical development was proposed8, based on the observation of similar behaviour of the two daughters of individual cells. These discrepancies in organ design emphasise the need for studies investigating individual cell progenies in other organ systems. Here we investigate how the allocation of endocrine and acinar fates is balanced with progenitor expansion from the beginning of pancreas formation using clonal analysis and singlecell molecular profiling. Embryonic mouse pancreas development is initiated at around embryonic day (E)9.0 by the specification of pancreatic progenitors at the dorsal and ventral sides of the posterior foregut endoderm9. Though induced by different mechanisms, the two anlage are composed of expanding Pdx1+Hnf1b+Sox9+Ptf1a+ progenitors forming bud-like structures protruding into the surrounding mesenchyme10. A small number of Neurog3+ endocrine precursors giving rise to the endocrine lineage of the pancreas are also found in these early buds11, 12. Morphogenetic processes occur concomitantly leading to the formation of lumens and their organisation into a plexus and subsequent tree-like branches13, 14. While the distal tip-domain is comprised of Ptf1a+ unipotent acinar progenitors after E13.515, 16, the Hnf1b+ trunk domain is bipotent and gives rise to endocrine cells, as well as the ductal cells that will eventually line the epithelial network draining acinar digestive enzymes to the duodenum17–19. Following specification towards the endocrine lineage, Neurog3+ endocrine precursors delaminate from the epithelial trunk domain to form immature islet clusters that will eventually mature into the endocrine Islets of Langerhans20. Although population-based lineage tracing has demonstrated the multipotency of the early pancreatic progenitors by virtue of their capability to give rise to progeny in the three major pancreatic lineages12, 15–17, 21 (Fig. 1a), no study has addressed the clonal contribution of the proposed multipotent pancreatic progenitors (MPCs) to pancreas organogenesis. One previous clonal analysis indeed restricted its focus on the progeny of single endocrine precursors examining their postnatal expansion22. Recent studies have demonstrated that pancreatic trunk progenitors undergo stochastic priming towards the endocrine lineage at mid-gestation19, and thus we questioned whether there are subpopulations of pancreatic progenitors exhibiting restricted lineage potencies from the onset of embryonic pancreas development or whether progeny from equipotent progenitors undergo stochastic lineage commitment. In this study, using clonal analysis of E9.5 pancreatic progenitors, when the pancreatic primordium has just been specified, we demonstrate that individual pancreatic progenitors contribute heterogeneously to pancreas organogenesis both in progeny size 2
and fate composition. While some progenitors are multipotent per se, giving rise to acinar, endocrine and ductal progeny, we also demonstrate the existence of bipotent ducto-endocrine and unipotent endocrine cells forming half of the primordium. This population represents cells at different stages of progression on the endocrine differentiation path, including proliferative endocrinecommitted cells, and exhibits undetectable to low levels of PTF1A. In contrast, bipotent and multipotent clones do not exhibit different expression profiles, suggesting they are not molecularly distinct cell populations. We show that clonal expansion and fate heterogeneity are compatible with a simple model of probabilistic cell fate acquisition operating downstream of spatially controlled proliferative and fate-biasing patterning cues. Results Single E9.5 pancreatic cells produce heterogeneous progeny. To investigate how individual cells among the about 500 cells that have just been specified to found the pancreas contribute to organogenesis, we devised a lineage tracing strategy making use of the Rosa26CreER driver (Fig. 1b). The ubiquitous activity of the Rosa26 locus ensures CreER expression throughout the developing embryo and hence also enables non-biased labelling of pancreatic cells23. We selected the mT/mG24 reporter over other multicolour reporters to be able to mark the differentiation status of clonal progeny. This required the dosage of very low levels of the active tamoxifen metabolite 4-OH tamoxifen (4-OHTm) to reach labelling of only one cell per pancreatic primordium within the 24 h following injection25. The labelling index of 11.8% (20 epithelial clones in 170 embryos) ensured a low risk of labelling two progenitors in the same pancreatic bud as of 1.4% (0.118×0.118). Whole-mount staining of E14.5 pancreata for endocrine (PAX6), progenitors lining the ducts (SOX9) and acinar (CPA1) markers enabled us to determine the fate of labelled GFP+ progeny at E14.5, a stage by which acinar cells are committed (Fig. 1c–h)12, 15–17. We observed a large extent of clone size heterogeneity, ranging from single-cell clones to clones consisting of hundreds of cells (Fig. 1h). Single GFP+ cells belonged to the endocrine lineage based on immunoreactivity for PAX6, cell morphology and location outside the pancreatic epithelium in islet-like structures (Fig. 1d, f, g). These single cells are expected to result from labelling non-proliferative endocrine cells, their Neurog3-expressing precursors or pancreatic progenitors differentiating directly into the endocrine lineage without dividing. We also observed 2- and 3-cell endocrine clones, suggesting that a labelled endocrine-biased progenitor had undergone a single or two rounds of divisions. In line with the postulated existence of multipotent progenitors based on nonclonal analyses, multipotent clones of 40–250 cells were found, consisting of ductal, endocrine and acinar progeny (Fig. 1h). Moreover we did observe bipotent clones of 6–100 cells harbouring only ductal and endocrine progeny, indicating that not all E9.5 progenitors contribute to the acinar lineage during pancreas organogenesis. However, we did not observe unipotent acinar clones arising from E9.5 progenitors. While confirming the existence of MPCs at the single-cell level, our results reveal heterogeneity in potency and contribution to pancreas organogenesis from single pancreatic cells. Furthermore they uncover the existence of bipotent progenitors as early as E9.5 and that half of the cells in the early pancreatic anlage give rise to solely endocrine progeny, a surprising finding considering that the adult endocrine cells only account for about 1% of the adult organ26. Heterogeneous marker expression in E9.5 progenitors. Heterogeneity in the clonal progeny may be either due to an intrinsic lineage bias in sub-populations of E9.5 progenitors or
NATURE COMMUNICATIONS | 8: 605
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
ARTICLE
a
Endocrine precursor NEUROG3
Pancreatic progenitors
PAX6 CHRA
Endocrine cells
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
NEUROG3 Duct
Trunk
HNF1B SOX9 Tip
HNF1B SOX9
PTF1A CPA
PTF1A CPA
Acinus
PDX1 SOX9 PTF1A HNF1B
b
E9.5
4-OHTm injection at E9.5 in mouse carrying Rosa26CreER;mT/mG embryos
Isolate E14.5 pancreas
Whole-mount immunostaining
Identify fates of clonal progeny
E13.5
g
c GFP SOX9 CPA1
d
GFP SOX9 CPA1
GFP
GFP SOX9
GFP SOX9 E-Cad
GFP SOX9
GFP PAX6
h
Rosa26CreER tracing E9.5 to E14.5 300 250
GFP SOX9
GFP 1 2
3
200
Cell number
e
150
Cell number
6 5 4 3 2 1
100
0 1 2 3 4 5 6 7 8 9 101112 Clone
50
f
GFP SOX9 PAX6
GFP SOX9 PAX6
GFP SOX9
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Clone Progenitors 1
2
Endocrine
Acinar
3
Fig. 1 Rosa26CreER-mediated clonal analysis reveals heterogeneous contribution of E9.5 progenitors to pancreas organogenesis. a Schematic overview of lineage relationships based on previous global lineage tracing. b Schematic overview of strategy applied to identify fates of clonal progeny from E9.5 pancreatic progenitors. c 3D maximum intensity projection (MIP) of a large, multipotent clone containing acinar (CPA1), progenitors lining the ducts (SOX9) and endocrine progeny (PAX6, from other multipotent clone) d. Scale bars, 100 µm c and 15 µm d. e 3D MIP of a bipotent clone containing endocrine (insets 1 and 2) and ductal (inset 3) clonal progeny f. Scale bars, 100 µm e and 10 µm f. g 3D MIP and optical section (inset) showing a single-labelled endocrine cell after clonal analysis from E9.5 to E14.5. Note the localisation of the GFP+ cell in an E-CADLow endocrine cluster. Scale bars, 80 µm and 15 µm (inset). h Quantification of clone sizes and compositions following clonal analysis from E9.5 to E14.5 (n = 170 embryos, 20 with clones, 1 excluded due to poor immunocytochemistry-the images displayed show representative data from those)
due to the stochastic fate allocation of clonal progeny during progressive organisation and compartmentalisation of the pancreatic epithelium. To test the first hypothesis, we conducted single-cell qRT-PCR following FACS isolation of dorsal foregut progenitor cells at E9.5 (Fig. 2a). tSNE-mediated dimensionality reduction of single cells revealed the existence of three distinct populations (Fig. 2b). On the basis of the expression of known lineage markers, we classify these three clusters as pancreatic endocrine (Neurog3+ and Glucagon+), pancreatic progenitors (Pdx1+ and Sox9+) and duodenal progenitors (Cdx2+, absence of Pdx1 and Sox9). Interestingly, cells characterised as belonging to the endocrine lineage organised on one projection axis forming a pseudo-temporal trajectory starting from Neurog3+ endocrine NATURE COMMUNICATIONS | 8: 605
precursors and progressing with the expression of markers associated with progressive endocrine maturation (Fig. 2c). These molecularly distinct cells are expected to contribute to the non-proliferative endocrine-committed cells observed in the lineage tracing. When focusing the dimensionality reduction on Pdx1+ pancreatic progenitors only, we observed marked heterogeneities in expression of various pancreas-associated transcription factors. Since at E14.5 Ptf1a marks acinar cells at the tip while Nkx6-116, 27, Hnf1b17 and Hes118 mark bipotent progenitors in the trunk, we investigated whether cells expressing these markers at E9.5 already had specific molecular signatures suggestive of emerging tip and trunk fates. However, in spite of heterogeneous expression of these markers, no global gene
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
3
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
FACS GFP+ single cells
Dissociate E9.5 Hes1-eGFP gut tube
1 2 3 4 5 6 7 8 9 10 11 12 A B C D E F G H
Melting
Amplification 0.18
Singe cell qPCR
0.15
0.10 0.08
–dRN / dT
Dorsal bud
ΔRN
a
0.06 0.04
0.12 0.09 0.06 0.03 0.00
0.02
–0.03
0.00 1
11
21
61
71
Cycle
81 Temperature
91
Ventral bud
c
b 10
Pdx1
Sox9
Cdx2
Hnf1b
Ptf1a
Neurog3
NeuroD
Glucagon
tSNE 2
5
0
–5
–10 –10
–5
0
5
10
tSNE 1 Mid
Low
e
d 4
High
Pdx1
Sox9
Hnf1b
Ptf1a
Nkx6.1
Dll1
Notch1
Hes1
3 2 tSNE 2
1 0 –1 –2 –3 –4 –4
–3
–2
–1
0
1
2
3
4
5
tSNE 1
Fig. 2 Single-cell qRT-PCR of E9.5 pancreatic progenitors reveals endocrine differentiation progression but no apparent tip-trunk segregation signature. a Schematic outline of the strategy used to isolate single pancreatic progenitor cells for single-cell qRT-PCR analysis using the Hes1-eGFP strain43. b tSNE dimensionality reduction of all single cells passing initial quality controls. Three cell clusters are displayed in green, black and red, corresponding to pancreatic progenitors, duodenal progenitors and pancreatic endocrine cells, respectively. c Hnf1b is expressed in both pancreatic and duodenal progenitors, whereas Ptf1a expression is exclusively detected in pancreatic progenitors albeit in heterogeneous pattern. Cells in the endocrine population cluster organise on a pseudo-temporal differentiation pathway starting with Neurog3+ endocrine precursors at the left, progressing into glucagon-expressing terminally differentiated endocrine cells at the right. d tSNE dimensionality reduction of Pdx1+ pancreatic progenitors only. e Despite heterogeneous expression of transcription factors and signalling components involved in subsequent tip-trunk segregation, expression of these markers seems uncorrelated and does not partition Pdx1+ progenitors into tip- and trunk-biased clusters
signature was associated with Nkx6-1, Hnf1b or Hes1, and these three markers showed no cross-correlation (Fig. 2d, e; Supplementary Figs. 1–4). Although Ptf1a expression did not correlate strongly with specific single markers, Ptf1a+-cells clustered after tSNE-mediated dimensionality reduction, suggesting that they are more similar to each other than to other progenitor cells. This molecular analysis uncovers that cells committed to endocrine differentiation can be molecularly identified, whereas subpopulations of multipotent or bipotent progenitors identified by clonal analysis cannot be molecularly predicted with this set of markers.
and measurements of the correlation in expression levels between neighbouring cells, we observed that HNF1B is expressed at higher levels towards the more posterior side of the pancreatic bud and in the gut tube and that PTF1A expressing cells appear clustered at a medial-bilateral location in E9.5 dorsal buds. Other transcription factors did not show any regionalisation of expression levels (Fig. 3d). These results demonstrate that the transcriptional profiles observed by single-cell qRT-PCR are translated into similar global profiles at the protein level, and that the levels of PTF1A and HNF1B display regionalised patterns in the E9.5 bud.
Spatial patterns of progenitor marker expression. To further assess heterogeneity in markers at the protein level, we used whole-mount immunostaining of E9.5 gut tubes and quantitative image analysis (Fig. 3a, b). We observed heterogeneous expression levels of HES1, SOX9 and PTF1A, whereas HNF1B and PDX1 were more homogeneously expressed among progenitors (Fig. 3c, d). Using 3D Voronoi-Delaunay triangulation (Fig. 3a)
Distinct clonal progeny of Hnf1b- and Ptf1a-expressing cells. The differential expression of PTF1A and HNF1B at E9.5 and the subsequent segregation of these markers to the tip and trunk domain, respectively, led us to investigate whether single progenitors expressing Ptf1a or Hnf1b at E9.5 contribute differential progeny by clonal analysis using Ptf1aCreER- and Hnf1bCreER drivers (Fig. 4a). The Hnf1bCreER-driver (labelling
4
NATURE COMMUNICATIONS | 8: 605
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
a
b
Voronoi cells
1. All cells, 3D distribution
Delaunay triangulation
Delaunay triangulation pruning by distance
4. 10 μm neighbour threshold
3. 5 μm neighbour threshold
2. Neighbours connected, no thresholding
= cell without neighbour
c Merge
HNF1B
SOX9
Merge
Intensity
Neighbour CV
Neighbour mean
HES1
SOX9
SOX9
d HES1
= cell with neighbour
PDX1
SOX9
Merge
HNF1B
D A
P
PTF1A
SOX9
Merge
PTF1A
PDX1
V
Low
Mid
High
Low
Mid
High
Fig. 3 E9.5 pancreatic progenitors display heterogeneous expression of pancreas-associated transcription factors and distinct patterns of neighbour cell expression correlation. a Schematic of Voronoi-Delaunay triangulation implemented to identify neighbour cells in 3D. Distance-based pruning was implemented in order to identify biologically meaningful neighbours in 3D space. b Example of Voronoi-Delaunay triangulation-mediated neighbour identification in E9.5 pancreatic bud. Cells colour-coded according to z-position (1) are subjected to Voronoi-Delaunay triangulation, identifying all neighbour relations in 3D (2). Implementation of a distance threshold of 5 μm (3) or 10 μm (4) generates different number of cells with and without 3D neighbours (green and red cells, respectively). For all downstream analyses, 10 μm was chosen as neighbour distance threshold. c Example of transcription factor expression pattern in E9.5 pancreatic buds following whole-mount staining. 3D MIP is displayed. Scale bars, 30 µm. d 3D plots of staining intensity, neighbour coefficient of variation and neighbour mean intensities from immunostaining against the indicated transcription factors. Note the heterogeneous expression patterns of HES1, SOX9 and PTF1A and the regionalised expression of HNF1B (posterior) and PTF1A (lateral) (n = 2 for all, except n = 8 for SOX9. The images displayed show representative data from those)
index: 35 clones in 120 pancreata, 29%; probability of double labelling, 8%) resulted in a similar pattern of clones as observed with Rosa26CreER, that is unipotent endocrine, bipotent ductoendocrine as well as multipotent clones (Fig. 4b, c and f). However, the endocrine-committed clones were less frequent, constituting 33% instead of 50% of the total clone repertoire, likely due to the fact that unlike Rosa26, Hnf1b is not expressed in mature endocrine cells28. In addition, we detect HNF1B immunoreactivity in 67.7 ± 3.8% of the NEUROG3-expressing endocrine precursors at this stage, while Rosa26 is expected to be expressed in all (Supplementary Fig. 5). A similar frequency of endocrine-committed precursors was observed when tracing Hnf1bCreER-labelled cells from E9.5 to E10.5 (Supplementary Fig. 6). Interestingly, we observed a clone consisting solely of 6 endocrine cells. Combined with the two 3-cell clones seen in the Rosa26 tracing, this suggests that some endocrinebiased progenitors can undergo multiple rounds of divisions (Fig. 4b, f; clone # 12), in line with the recent observation that cells with low levels of Neurog3 transcription can proliferate29. The low-differentiation rate towards the endocrine lineage (p = 0.12 ± 0.007, Supplementary Fig. 7) makes independent probabilistic entry of progeny into the endocrine lineage highly NATURE COMMUNICATIONS | 8: 605
unlikely, suggesting that the E9.5 pancreatic bud contains progenitors biased towards multigenerational endocrine specification. On the other hand, we never observed unipotent acinar clones arising from E9.5 progenitors. Furthermore, acinarcontaining clones always contained ductal and endocrine progeny, suggesting that acinar-lineage allocation has not yet occurred in any cell at E9.5. The proportion of multipotent clones was similar to what was observed using the Rosa26CreER driver after correction for the absence of labelling of mature endocrine cells by Hnf1bCreER. Though only five of the clones were found in the ventral pancreas, which is smaller than the dorsal, they were bi- or multipotent but not endocrine committed, possibly due to a delay in endocrine program onset in the ventral pancreas. This suggests that the assumed labelling of progenitors with high-expression levels of Hnf1b expression at E9.5 does not bias lineage contribution to the trunk domain (Fig. 4f). Similarly, Ptf1aCreER-based lineage tracing did not bias progenitors towards the acinar lineage, either (Fig. 4d, e and g). However, cells traced by Ptf1aCreER (labelling index: 13 clones in 30 dorsal pancreata, 44%; probability of double labelling, 19%) did not form endocrine-only clones, unlike what was seen with Rosa26CreERand Hnf1bCreER-drivers. This suggests that cells exhibiting high
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
5
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
a
Isolate E14.5 pancreas
4-OHTm injection at E9.5 in mouse carrying Hnf1bCreER;mT/mG or Ptf1aCreER;mT/mG embryos
b
Hnf1bCreER E9.5 to E14.5 GFP
GFP SOX9 CPA1 PAX6
1
Identify fates of clonal progeny
Whole-mount immunostaining
f
Hnf1bCreER E9.5 to E14.5 600
2
15 500
10
c
5 300
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Cell number
400
200
1
SOX9 PAX6
GFP SOX9 PAX6
2
GFP SOX9 PAX6
SOX9 PAX6 100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
0 Clone
d
Ptf1aCreER E9.5 to E14.5
Progenitor
Acinar
Endocrine
GFP
GFP SOX9 CPA1
g
Ptf1aCreER E9.5 to E14.5 50
500
40 30
e GFP CPA1
GFP SOX9
GFP SOX9 PAX6
Cell number
400
20 10
300
0 1
2
2
3
3
4
5
6
7
200 100 0 1
4
5
6
7 8 Clone
9
10 11 12 13
Fig. 4 Hnf1bCreER- and Ptf1aCreER-mediated clonal analysis confirms the existence of ducto-endocrine bipotent clones and the absence of unipotent acinar progenitors at E9.5. a Schematic overview of strategy applied to identify fates of clonal progeny from E9.5 pancreatic progenitors. b 3D MIP and high-magnification optical sections (c) of clonal progeny from a unipotent endocrine clone containing six labelled endocrine PAX6+ endocrine cells generated by Hnf1bCreER-mediated clonal analysis. Arrow indicates two juxtaposed GFP+ cells, whereas arrowheads indicate single GFP-labelled cells. Scale bars, 100 µm b and 15 µm c. d 3D MIP and high-magnification optical sections (e) of a multipotent clone derived from Ptf1aCreER-mediated lineage tracing. Scale bars, 150 µm d and 15 µm e. f, g Quantification of clone sizes and fate compositions following Hnf1bCreER- and Ptf1aCreER-mediated clonal analysis, respectively (n = 120 embryos for Hnf1bCreER and 34 embryos for Ptf1aCreER- the images displayed show representative data from those). Note that clone no. 1 contains only one cell and is only SOX9+. Most clones were found in the dorsal pancreas, except clones no. 14, 17, 18, 26, 33 in f and clones no. 3, 8, 9 in g which were found in the ventral pancreas
Ptf1a expression at around E9.5 do not immediately form endocrine cells, unlike progenitors traced by Rosa26CreER and Hnf1bCreER, though they retain endocrine differentiation capacity as these cells give rise to endocrine-containing clones later in their clonal evolution. This hypothesis is supported by Ptf1a anticorrelation with early markers of endocrine differentiation such as Mfng and Neurog3 in our single-cell qRT-PCR analysis at E9.5 (Supplementary Fig. 3). 6
A probabilistic model of progenitor progeny fate allocation. The apparent lack of tip-trunk biased progenitors suggested by both single-cell analysis and tracing at E9.5 led us to investigate whether a model of probabilistic cell-fate choices could recapitulate the in vivo clonal distribution data. To this end, we constructed a mathematical model of in silico clonal growth by simulating cell divisions over a period spanning the in vivo clonal tracing. Every time a cell gave rise to progeny through cell division, clonal progeny were fate-allocated with a probability of
NATURE COMMUNICATIONS | 8: 605
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
Model 2 f ~ cos(t )
Model 1 f = const
a 1.0 0.8
c⋅(f )
0.6 0.4 0.2
Probability
Fraction
1–c
0.6 0.4 0.2
q
q
0.0
5
0.0 14.5 13.5 12.5 11.5 10.5 9.5
14.5 13.5 12.5 11.5 10.5 9.5
Progenitors
0.30 0.25 0.20 0.15 0.10 0.05 0.00
0.8 Fraction
c⋅(1–f )
Cell cycle length distribution
1.0
f (t) 1-f (t )
Time (days)
Time (days)
Endocrine
10 15 20 25 30 35 Time (h)
Acinar
Simulations for one parameter set (c,q)
c
Fractions of cell types averaged over 100 clones 1.0 0.8 0.6 0.4 0.2 0.0
x 25
5
5
4.
5
2.
E1
5
1.
E1
0.
5
E1
5
4.
E1
5
2.
E1
E1
5
1.
5
0.
5
4.
d
E1
5
2.
E1
5
1.
E1
x 2500
E1
E1
0.
...
E1
Examples of the in silico clones
Frequency
b
Clonal composition E9.5 –E14.5
Clonal size, [cells]
... 160 140 120 100 80 60 40 20 0
x 2500 ...
0
20
10
30
2490
2500
Clone
e
Parameter scan Model 2 f ~ cos(t )
f
0.95 10
OPEN
Stochastic priming and spatial cues orchestrate heterogeneous clonal contribution to mouse pancreas organogenesis Hjalte List Larsen1, Laura Martín-Coll1, Alexander Valentin Nielsen2, Christopher V. E. Wright3, Ala Trusina2, Yung Hae Kim1 & Anne Grapin-Botton 1
Spatiotemporal balancing of cellular proliferation and differentiation is crucial for postnatal tissue homoeostasis and organogenesis. During embryonic development, pancreatic progenitors simultaneously proliferate and differentiate into the endocrine, ductal and acinar lineages. Using in vivo clonal analysis in the founder population of the pancreas here we reveal highly heterogeneous contribution of single progenitors to organ formation. While some progenitors are bona fide multipotent and contribute progeny to all major pancreatic cell lineages, we also identify numerous unipotent endocrine and ducto-endocrine bipotent clones. Single-cell transcriptional profiling at E9.5 reveals that endocrine-committed cells are molecularly distinct, whereas multipotent and bipotent progenitors do not exhibit different expression profiles. Clone size and composition support a probabilistic model of cell fate allocation and in silico simulations predict a transient wave of acinar differentiation around E11.5, while endocrine differentiation is proportionally decreased. Increased proliferative capacity of outer progenitors is further proposed to impact clonal expansion.
1 DanStem, University of Copenhagen, 3B Blegdamsvej, DK-2200 Copenhagen N, Denmark. 2 Niels Bohr Institute, University of Copenhagen, 17 Blegdamsvej, DK-2200 Copenhagen N, Denmark. 3 Department of Cell & Developmental Biology, Vanderbilt University, Nashville, TN 37232-0494, USA. Correspondence and requests for materials should be addressed to Y.H.K. (email: [email protected]) or to A.G.-B. (email: [email protected])
NATURE COMMUNICATIONS | 8: 605
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
1
ARTICLE
D
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
efining the rules governing embryonic organ development and postnatal tissue homoeostasis is essential for understanding disease pathology and for the generation of functional cell types for regenerative medicine purposes. Seminal studies have demonstrated how rapidly proliferating postnatal tissues such as the skin and the intestine are homeostatically maintained by equipotent stem cells undergoing seemingly stochastic cell fates choices by neutral competition for limited niche signals1–4. In contrast to postnatal tissue homoeostasis, embryonic development of most organs occurs at a state of system disequilibrium, as a population of progenitors expands while simultaneously giving rise to differentiating progeny. Although optimality in the design of strategies ensuring rapid organ development has been proposed5, little is known regarding how global embryonic organogenesis is orchestrated when deconstructed into clonal units originating from single progenitors at the onset of organ bud formation. Studies of retinal development have provided compelling evidence for a stochastic process of cell fate choices using both in vitro6 and in vivo approaches7. However, a deterministic model of embryonic neocortical development was proposed8, based on the observation of similar behaviour of the two daughters of individual cells. These discrepancies in organ design emphasise the need for studies investigating individual cell progenies in other organ systems. Here we investigate how the allocation of endocrine and acinar fates is balanced with progenitor expansion from the beginning of pancreas formation using clonal analysis and singlecell molecular profiling. Embryonic mouse pancreas development is initiated at around embryonic day (E)9.0 by the specification of pancreatic progenitors at the dorsal and ventral sides of the posterior foregut endoderm9. Though induced by different mechanisms, the two anlage are composed of expanding Pdx1+Hnf1b+Sox9+Ptf1a+ progenitors forming bud-like structures protruding into the surrounding mesenchyme10. A small number of Neurog3+ endocrine precursors giving rise to the endocrine lineage of the pancreas are also found in these early buds11, 12. Morphogenetic processes occur concomitantly leading to the formation of lumens and their organisation into a plexus and subsequent tree-like branches13, 14. While the distal tip-domain is comprised of Ptf1a+ unipotent acinar progenitors after E13.515, 16, the Hnf1b+ trunk domain is bipotent and gives rise to endocrine cells, as well as the ductal cells that will eventually line the epithelial network draining acinar digestive enzymes to the duodenum17–19. Following specification towards the endocrine lineage, Neurog3+ endocrine precursors delaminate from the epithelial trunk domain to form immature islet clusters that will eventually mature into the endocrine Islets of Langerhans20. Although population-based lineage tracing has demonstrated the multipotency of the early pancreatic progenitors by virtue of their capability to give rise to progeny in the three major pancreatic lineages12, 15–17, 21 (Fig. 1a), no study has addressed the clonal contribution of the proposed multipotent pancreatic progenitors (MPCs) to pancreas organogenesis. One previous clonal analysis indeed restricted its focus on the progeny of single endocrine precursors examining their postnatal expansion22. Recent studies have demonstrated that pancreatic trunk progenitors undergo stochastic priming towards the endocrine lineage at mid-gestation19, and thus we questioned whether there are subpopulations of pancreatic progenitors exhibiting restricted lineage potencies from the onset of embryonic pancreas development or whether progeny from equipotent progenitors undergo stochastic lineage commitment. In this study, using clonal analysis of E9.5 pancreatic progenitors, when the pancreatic primordium has just been specified, we demonstrate that individual pancreatic progenitors contribute heterogeneously to pancreas organogenesis both in progeny size 2
and fate composition. While some progenitors are multipotent per se, giving rise to acinar, endocrine and ductal progeny, we also demonstrate the existence of bipotent ducto-endocrine and unipotent endocrine cells forming half of the primordium. This population represents cells at different stages of progression on the endocrine differentiation path, including proliferative endocrinecommitted cells, and exhibits undetectable to low levels of PTF1A. In contrast, bipotent and multipotent clones do not exhibit different expression profiles, suggesting they are not molecularly distinct cell populations. We show that clonal expansion and fate heterogeneity are compatible with a simple model of probabilistic cell fate acquisition operating downstream of spatially controlled proliferative and fate-biasing patterning cues. Results Single E9.5 pancreatic cells produce heterogeneous progeny. To investigate how individual cells among the about 500 cells that have just been specified to found the pancreas contribute to organogenesis, we devised a lineage tracing strategy making use of the Rosa26CreER driver (Fig. 1b). The ubiquitous activity of the Rosa26 locus ensures CreER expression throughout the developing embryo and hence also enables non-biased labelling of pancreatic cells23. We selected the mT/mG24 reporter over other multicolour reporters to be able to mark the differentiation status of clonal progeny. This required the dosage of very low levels of the active tamoxifen metabolite 4-OH tamoxifen (4-OHTm) to reach labelling of only one cell per pancreatic primordium within the 24 h following injection25. The labelling index of 11.8% (20 epithelial clones in 170 embryos) ensured a low risk of labelling two progenitors in the same pancreatic bud as of 1.4% (0.118×0.118). Whole-mount staining of E14.5 pancreata for endocrine (PAX6), progenitors lining the ducts (SOX9) and acinar (CPA1) markers enabled us to determine the fate of labelled GFP+ progeny at E14.5, a stage by which acinar cells are committed (Fig. 1c–h)12, 15–17. We observed a large extent of clone size heterogeneity, ranging from single-cell clones to clones consisting of hundreds of cells (Fig. 1h). Single GFP+ cells belonged to the endocrine lineage based on immunoreactivity for PAX6, cell morphology and location outside the pancreatic epithelium in islet-like structures (Fig. 1d, f, g). These single cells are expected to result from labelling non-proliferative endocrine cells, their Neurog3-expressing precursors or pancreatic progenitors differentiating directly into the endocrine lineage without dividing. We also observed 2- and 3-cell endocrine clones, suggesting that a labelled endocrine-biased progenitor had undergone a single or two rounds of divisions. In line with the postulated existence of multipotent progenitors based on nonclonal analyses, multipotent clones of 40–250 cells were found, consisting of ductal, endocrine and acinar progeny (Fig. 1h). Moreover we did observe bipotent clones of 6–100 cells harbouring only ductal and endocrine progeny, indicating that not all E9.5 progenitors contribute to the acinar lineage during pancreas organogenesis. However, we did not observe unipotent acinar clones arising from E9.5 progenitors. While confirming the existence of MPCs at the single-cell level, our results reveal heterogeneity in potency and contribution to pancreas organogenesis from single pancreatic cells. Furthermore they uncover the existence of bipotent progenitors as early as E9.5 and that half of the cells in the early pancreatic anlage give rise to solely endocrine progeny, a surprising finding considering that the adult endocrine cells only account for about 1% of the adult organ26. Heterogeneous marker expression in E9.5 progenitors. Heterogeneity in the clonal progeny may be either due to an intrinsic lineage bias in sub-populations of E9.5 progenitors or
NATURE COMMUNICATIONS | 8: 605
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
ARTICLE
a
Endocrine precursor NEUROG3
Pancreatic progenitors
PAX6 CHRA
Endocrine cells
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
NEUROG3 Duct
Trunk
HNF1B SOX9 Tip
HNF1B SOX9
PTF1A CPA
PTF1A CPA
Acinus
PDX1 SOX9 PTF1A HNF1B
b
E9.5
4-OHTm injection at E9.5 in mouse carrying Rosa26CreER;mT/mG embryos
Isolate E14.5 pancreas
Whole-mount immunostaining
Identify fates of clonal progeny
E13.5
g
c GFP SOX9 CPA1
d
GFP SOX9 CPA1
GFP
GFP SOX9
GFP SOX9 E-Cad
GFP SOX9
GFP PAX6
h
Rosa26CreER tracing E9.5 to E14.5 300 250
GFP SOX9
GFP 1 2
3
200
Cell number
e
150
Cell number
6 5 4 3 2 1
100
0 1 2 3 4 5 6 7 8 9 101112 Clone
50
f
GFP SOX9 PAX6
GFP SOX9 PAX6
GFP SOX9
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Clone Progenitors 1
2
Endocrine
Acinar
3
Fig. 1 Rosa26CreER-mediated clonal analysis reveals heterogeneous contribution of E9.5 progenitors to pancreas organogenesis. a Schematic overview of lineage relationships based on previous global lineage tracing. b Schematic overview of strategy applied to identify fates of clonal progeny from E9.5 pancreatic progenitors. c 3D maximum intensity projection (MIP) of a large, multipotent clone containing acinar (CPA1), progenitors lining the ducts (SOX9) and endocrine progeny (PAX6, from other multipotent clone) d. Scale bars, 100 µm c and 15 µm d. e 3D MIP of a bipotent clone containing endocrine (insets 1 and 2) and ductal (inset 3) clonal progeny f. Scale bars, 100 µm e and 10 µm f. g 3D MIP and optical section (inset) showing a single-labelled endocrine cell after clonal analysis from E9.5 to E14.5. Note the localisation of the GFP+ cell in an E-CADLow endocrine cluster. Scale bars, 80 µm and 15 µm (inset). h Quantification of clone sizes and compositions following clonal analysis from E9.5 to E14.5 (n = 170 embryos, 20 with clones, 1 excluded due to poor immunocytochemistry-the images displayed show representative data from those)
due to the stochastic fate allocation of clonal progeny during progressive organisation and compartmentalisation of the pancreatic epithelium. To test the first hypothesis, we conducted single-cell qRT-PCR following FACS isolation of dorsal foregut progenitor cells at E9.5 (Fig. 2a). tSNE-mediated dimensionality reduction of single cells revealed the existence of three distinct populations (Fig. 2b). On the basis of the expression of known lineage markers, we classify these three clusters as pancreatic endocrine (Neurog3+ and Glucagon+), pancreatic progenitors (Pdx1+ and Sox9+) and duodenal progenitors (Cdx2+, absence of Pdx1 and Sox9). Interestingly, cells characterised as belonging to the endocrine lineage organised on one projection axis forming a pseudo-temporal trajectory starting from Neurog3+ endocrine NATURE COMMUNICATIONS | 8: 605
precursors and progressing with the expression of markers associated with progressive endocrine maturation (Fig. 2c). These molecularly distinct cells are expected to contribute to the non-proliferative endocrine-committed cells observed in the lineage tracing. When focusing the dimensionality reduction on Pdx1+ pancreatic progenitors only, we observed marked heterogeneities in expression of various pancreas-associated transcription factors. Since at E14.5 Ptf1a marks acinar cells at the tip while Nkx6-116, 27, Hnf1b17 and Hes118 mark bipotent progenitors in the trunk, we investigated whether cells expressing these markers at E9.5 already had specific molecular signatures suggestive of emerging tip and trunk fates. However, in spite of heterogeneous expression of these markers, no global gene
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
3
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
FACS GFP+ single cells
Dissociate E9.5 Hes1-eGFP gut tube
1 2 3 4 5 6 7 8 9 10 11 12 A B C D E F G H
Melting
Amplification 0.18
Singe cell qPCR
0.15
0.10 0.08
–dRN / dT
Dorsal bud
ΔRN
a
0.06 0.04
0.12 0.09 0.06 0.03 0.00
0.02
–0.03
0.00 1
11
21
61
71
Cycle
81 Temperature
91
Ventral bud
c
b 10
Pdx1
Sox9
Cdx2
Hnf1b
Ptf1a
Neurog3
NeuroD
Glucagon
tSNE 2
5
0
–5
–10 –10
–5
0
5
10
tSNE 1 Mid
Low
e
d 4
High
Pdx1
Sox9
Hnf1b
Ptf1a
Nkx6.1
Dll1
Notch1
Hes1
3 2 tSNE 2
1 0 –1 –2 –3 –4 –4
–3
–2
–1
0
1
2
3
4
5
tSNE 1
Fig. 2 Single-cell qRT-PCR of E9.5 pancreatic progenitors reveals endocrine differentiation progression but no apparent tip-trunk segregation signature. a Schematic outline of the strategy used to isolate single pancreatic progenitor cells for single-cell qRT-PCR analysis using the Hes1-eGFP strain43. b tSNE dimensionality reduction of all single cells passing initial quality controls. Three cell clusters are displayed in green, black and red, corresponding to pancreatic progenitors, duodenal progenitors and pancreatic endocrine cells, respectively. c Hnf1b is expressed in both pancreatic and duodenal progenitors, whereas Ptf1a expression is exclusively detected in pancreatic progenitors albeit in heterogeneous pattern. Cells in the endocrine population cluster organise on a pseudo-temporal differentiation pathway starting with Neurog3+ endocrine precursors at the left, progressing into glucagon-expressing terminally differentiated endocrine cells at the right. d tSNE dimensionality reduction of Pdx1+ pancreatic progenitors only. e Despite heterogeneous expression of transcription factors and signalling components involved in subsequent tip-trunk segregation, expression of these markers seems uncorrelated and does not partition Pdx1+ progenitors into tip- and trunk-biased clusters
signature was associated with Nkx6-1, Hnf1b or Hes1, and these three markers showed no cross-correlation (Fig. 2d, e; Supplementary Figs. 1–4). Although Ptf1a expression did not correlate strongly with specific single markers, Ptf1a+-cells clustered after tSNE-mediated dimensionality reduction, suggesting that they are more similar to each other than to other progenitor cells. This molecular analysis uncovers that cells committed to endocrine differentiation can be molecularly identified, whereas subpopulations of multipotent or bipotent progenitors identified by clonal analysis cannot be molecularly predicted with this set of markers.
and measurements of the correlation in expression levels between neighbouring cells, we observed that HNF1B is expressed at higher levels towards the more posterior side of the pancreatic bud and in the gut tube and that PTF1A expressing cells appear clustered at a medial-bilateral location in E9.5 dorsal buds. Other transcription factors did not show any regionalisation of expression levels (Fig. 3d). These results demonstrate that the transcriptional profiles observed by single-cell qRT-PCR are translated into similar global profiles at the protein level, and that the levels of PTF1A and HNF1B display regionalised patterns in the E9.5 bud.
Spatial patterns of progenitor marker expression. To further assess heterogeneity in markers at the protein level, we used whole-mount immunostaining of E9.5 gut tubes and quantitative image analysis (Fig. 3a, b). We observed heterogeneous expression levels of HES1, SOX9 and PTF1A, whereas HNF1B and PDX1 were more homogeneously expressed among progenitors (Fig. 3c, d). Using 3D Voronoi-Delaunay triangulation (Fig. 3a)
Distinct clonal progeny of Hnf1b- and Ptf1a-expressing cells. The differential expression of PTF1A and HNF1B at E9.5 and the subsequent segregation of these markers to the tip and trunk domain, respectively, led us to investigate whether single progenitors expressing Ptf1a or Hnf1b at E9.5 contribute differential progeny by clonal analysis using Ptf1aCreER- and Hnf1bCreER drivers (Fig. 4a). The Hnf1bCreER-driver (labelling
4
NATURE COMMUNICATIONS | 8: 605
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
a
b
Voronoi cells
1. All cells, 3D distribution
Delaunay triangulation
Delaunay triangulation pruning by distance
4. 10 μm neighbour threshold
3. 5 μm neighbour threshold
2. Neighbours connected, no thresholding
= cell without neighbour
c Merge
HNF1B
SOX9
Merge
Intensity
Neighbour CV
Neighbour mean
HES1
SOX9
SOX9
d HES1
= cell with neighbour
PDX1
SOX9
Merge
HNF1B
D A
P
PTF1A
SOX9
Merge
PTF1A
PDX1
V
Low
Mid
High
Low
Mid
High
Fig. 3 E9.5 pancreatic progenitors display heterogeneous expression of pancreas-associated transcription factors and distinct patterns of neighbour cell expression correlation. a Schematic of Voronoi-Delaunay triangulation implemented to identify neighbour cells in 3D. Distance-based pruning was implemented in order to identify biologically meaningful neighbours in 3D space. b Example of Voronoi-Delaunay triangulation-mediated neighbour identification in E9.5 pancreatic bud. Cells colour-coded according to z-position (1) are subjected to Voronoi-Delaunay triangulation, identifying all neighbour relations in 3D (2). Implementation of a distance threshold of 5 μm (3) or 10 μm (4) generates different number of cells with and without 3D neighbours (green and red cells, respectively). For all downstream analyses, 10 μm was chosen as neighbour distance threshold. c Example of transcription factor expression pattern in E9.5 pancreatic buds following whole-mount staining. 3D MIP is displayed. Scale bars, 30 µm. d 3D plots of staining intensity, neighbour coefficient of variation and neighbour mean intensities from immunostaining against the indicated transcription factors. Note the heterogeneous expression patterns of HES1, SOX9 and PTF1A and the regionalised expression of HNF1B (posterior) and PTF1A (lateral) (n = 2 for all, except n = 8 for SOX9. The images displayed show representative data from those)
index: 35 clones in 120 pancreata, 29%; probability of double labelling, 8%) resulted in a similar pattern of clones as observed with Rosa26CreER, that is unipotent endocrine, bipotent ductoendocrine as well as multipotent clones (Fig. 4b, c and f). However, the endocrine-committed clones were less frequent, constituting 33% instead of 50% of the total clone repertoire, likely due to the fact that unlike Rosa26, Hnf1b is not expressed in mature endocrine cells28. In addition, we detect HNF1B immunoreactivity in 67.7 ± 3.8% of the NEUROG3-expressing endocrine precursors at this stage, while Rosa26 is expected to be expressed in all (Supplementary Fig. 5). A similar frequency of endocrine-committed precursors was observed when tracing Hnf1bCreER-labelled cells from E9.5 to E10.5 (Supplementary Fig. 6). Interestingly, we observed a clone consisting solely of 6 endocrine cells. Combined with the two 3-cell clones seen in the Rosa26 tracing, this suggests that some endocrinebiased progenitors can undergo multiple rounds of divisions (Fig. 4b, f; clone # 12), in line with the recent observation that cells with low levels of Neurog3 transcription can proliferate29. The low-differentiation rate towards the endocrine lineage (p = 0.12 ± 0.007, Supplementary Fig. 7) makes independent probabilistic entry of progeny into the endocrine lineage highly NATURE COMMUNICATIONS | 8: 605
unlikely, suggesting that the E9.5 pancreatic bud contains progenitors biased towards multigenerational endocrine specification. On the other hand, we never observed unipotent acinar clones arising from E9.5 progenitors. Furthermore, acinarcontaining clones always contained ductal and endocrine progeny, suggesting that acinar-lineage allocation has not yet occurred in any cell at E9.5. The proportion of multipotent clones was similar to what was observed using the Rosa26CreER driver after correction for the absence of labelling of mature endocrine cells by Hnf1bCreER. Though only five of the clones were found in the ventral pancreas, which is smaller than the dorsal, they were bi- or multipotent but not endocrine committed, possibly due to a delay in endocrine program onset in the ventral pancreas. This suggests that the assumed labelling of progenitors with high-expression levels of Hnf1b expression at E9.5 does not bias lineage contribution to the trunk domain (Fig. 4f). Similarly, Ptf1aCreER-based lineage tracing did not bias progenitors towards the acinar lineage, either (Fig. 4d, e and g). However, cells traced by Ptf1aCreER (labelling index: 13 clones in 30 dorsal pancreata, 44%; probability of double labelling, 19%) did not form endocrine-only clones, unlike what was seen with Rosa26CreERand Hnf1bCreER-drivers. This suggests that cells exhibiting high
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
5
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
a
Isolate E14.5 pancreas
4-OHTm injection at E9.5 in mouse carrying Hnf1bCreER;mT/mG or Ptf1aCreER;mT/mG embryos
b
Hnf1bCreER E9.5 to E14.5 GFP
GFP SOX9 CPA1 PAX6
1
Identify fates of clonal progeny
Whole-mount immunostaining
f
Hnf1bCreER E9.5 to E14.5 600
2
15 500
10
c
5 300
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Cell number
400
200
1
SOX9 PAX6
GFP SOX9 PAX6
2
GFP SOX9 PAX6
SOX9 PAX6 100
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
0 Clone
d
Ptf1aCreER E9.5 to E14.5
Progenitor
Acinar
Endocrine
GFP
GFP SOX9 CPA1
g
Ptf1aCreER E9.5 to E14.5 50
500
40 30
e GFP CPA1
GFP SOX9
GFP SOX9 PAX6
Cell number
400
20 10
300
0 1
2
2
3
3
4
5
6
7
200 100 0 1
4
5
6
7 8 Clone
9
10 11 12 13
Fig. 4 Hnf1bCreER- and Ptf1aCreER-mediated clonal analysis confirms the existence of ducto-endocrine bipotent clones and the absence of unipotent acinar progenitors at E9.5. a Schematic overview of strategy applied to identify fates of clonal progeny from E9.5 pancreatic progenitors. b 3D MIP and high-magnification optical sections (c) of clonal progeny from a unipotent endocrine clone containing six labelled endocrine PAX6+ endocrine cells generated by Hnf1bCreER-mediated clonal analysis. Arrow indicates two juxtaposed GFP+ cells, whereas arrowheads indicate single GFP-labelled cells. Scale bars, 100 µm b and 15 µm c. d 3D MIP and high-magnification optical sections (e) of a multipotent clone derived from Ptf1aCreER-mediated lineage tracing. Scale bars, 150 µm d and 15 µm e. f, g Quantification of clone sizes and fate compositions following Hnf1bCreER- and Ptf1aCreER-mediated clonal analysis, respectively (n = 120 embryos for Hnf1bCreER and 34 embryos for Ptf1aCreER- the images displayed show representative data from those). Note that clone no. 1 contains only one cell and is only SOX9+. Most clones were found in the dorsal pancreas, except clones no. 14, 17, 18, 26, 33 in f and clones no. 3, 8, 9 in g which were found in the ventral pancreas
Ptf1a expression at around E9.5 do not immediately form endocrine cells, unlike progenitors traced by Rosa26CreER and Hnf1bCreER, though they retain endocrine differentiation capacity as these cells give rise to endocrine-containing clones later in their clonal evolution. This hypothesis is supported by Ptf1a anticorrelation with early markers of endocrine differentiation such as Mfng and Neurog3 in our single-cell qRT-PCR analysis at E9.5 (Supplementary Fig. 3). 6
A probabilistic model of progenitor progeny fate allocation. The apparent lack of tip-trunk biased progenitors suggested by both single-cell analysis and tracing at E9.5 led us to investigate whether a model of probabilistic cell-fate choices could recapitulate the in vivo clonal distribution data. To this end, we constructed a mathematical model of in silico clonal growth by simulating cell divisions over a period spanning the in vivo clonal tracing. Every time a cell gave rise to progeny through cell division, clonal progeny were fate-allocated with a probability of
NATURE COMMUNICATIONS | 8: 605
| DOI: 10.1038/s41467-017-00258-4 | www.nature.com/naturecommunications
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00258-4
Model 2 f ~ cos(t )
Model 1 f = const
a 1.0 0.8
c⋅(f )
0.6 0.4 0.2
Probability
Fraction
1–c
0.6 0.4 0.2
q
q
0.0
5
0.0 14.5 13.5 12.5 11.5 10.5 9.5
14.5 13.5 12.5 11.5 10.5 9.5
Progenitors
0.30 0.25 0.20 0.15 0.10 0.05 0.00
0.8 Fraction
c⋅(1–f )
Cell cycle length distribution
1.0
f (t) 1-f (t )
Time (days)
Time (days)
Endocrine
10 15 20 25 30 35 Time (h)
Acinar
Simulations for one parameter set (c,q)
c
Fractions of cell types averaged over 100 clones 1.0 0.8 0.6 0.4 0.2 0.0
x 25
5
5
4.
5
2.
E1
5
1.
E1
0.
5
E1
5
4.
E1
5
2.
E1
E1
5
1.
5
0.
5
4.
d
E1
5
2.
E1
5
1.
E1
x 2500
E1
E1
0.
...
E1
Examples of the in silico clones
Frequency
b
Clonal composition E9.5 –E14.5
Clonal size, [cells]
... 160 140 120 100 80 60 40 20 0
x 2500 ...
0
20
10
30
2490
2500
Clone
e
Parameter scan Model 2 f ~ cos(t )
f
0.95 10
